Deep Learning-Based Target Tracking Approach for Threaded Sleeve Production Line of Prefabricated Building

As a novel building forms, prefabricated buildings offer unique advantages in terms of safety and speed. Focusing on the problems of target tracking in the threaded sleeve production line of prefabricated buildings, such as low accuracy rate, slow recognition speed, and difficulty in dealing with object occlusion and environment changes, a target tracking and detection approach based on the improved DeepSort algorithm is proposed. The real-time tracking of the equipment of the prefabricated building threaded sleeve production line is realized. The improved YOLOv5 replaces the detection algorithm in the track phase of the original DeepSort algorithm, which is used for real-time tracking and detecting the equipment. The experimental results show that the tracking accuracy of the improved algorithm is 88%, 70.5%, and 64.8% for three videos of different lengths, respectively. The tracking detection frame rate reached 71.95 fps in the three experiments of video tracking. The improvement of the frame rate of the improved algorithm is over 100% compared with the traditional tracking algorithms. The improved algorithm enhances the accuracy of target tracking in prefabricated building threaded sleeve production line and improves the production efficiency of prefabricated building threaded sleeve production line effectively.